As for the servo, for example, RD50 (14 Nm, 80W, 440 g) - somewhere around 2.290 USD.This is already the price for a servo with a gearbox, a controller, 2 magnetic encoders and frameless motors.All components are built into the case itself.Large models are slightly more expensive.If unframed - then the price is about 330 USD.But, a lot depends on the party and the project.What are you working on?Mention!We are pleased.According to plans and markets: in April we release the first batch of serial 2nd generation cobots.This year we want to focus on the Russian-Belarusian segment.Satisfy accepted applications and pre-orders from all regions (Europe, Asia and Oceania, North America), but marketing and promotion for lead generation should be done, first of all, in the Russian Federation.According to recent IFR data, the density of robots in the Russian Federation is one of the lowest in the world - 1 unit per 10,000 workers.http://markets.businessinsider.com/news/stocks/SMBs-to-Drive-Collaborative-Robotics-Growth-as-they-Seek-a-More-Flexible-Manufacturing-Process-1015049546 Therefore, given the similarity of mentality, localization of production, market size, etc., we see a serious potential for ourselves in this region.As for non-production, cobots for fun are in demand, first of all, in cruise ships, banks, IT companies, conferences, chain hotels, etc. In this case, a cobot is such an eye-catcher.It slows down and localizes traffic passing by + a large number of shares in the network.And the issue of monetization is already a decision of a specific business.For many, by the way, this is a way to be "advanced" in the eyes of customers, partners, etc.I am sure that with the reduction in cost of cobots and their growth in popularity (a small number of people know about this), they will be more and more in demand in the non-industrial market.That is exactly what we want to get to.Imagine how great it is to load dishes into the dishwasher, clean up toys after children, etc. The lower the price drops, the programming becomes easier, the safer it increases - the closer this moment will be.By the way, the Moley Robotic Kitchens project is a prime example of this.https://www.youtube.com/watch?v=PE94ZA8E52MThe very moment when I am proud of our compatriots!Well done, I will recommend you to friends and acquaintances.In fact, exactly 2 years ago, I myself wanted to assemble a manipulator similar to yours at home, then I spent at least 2 full months on it in total over a period of 7-8 months, but I never achieved success due to of the problems you described, the main of which is the delivery time of 6 months, I’ll immediately make a reservation that I have no problems with money and I was ready to spend money on my toy, but I didn’t want to buy a ready-made manipulator because there was a desire to do everything myself.Of course, I didn’t get to the production of motors, where I stopped - brushless maxon motors and harmonic drive as a gearbox, because the task was to make not just a craft, but to make it no worse than Universal Robots, which I (and you, judging by the design of your manipulator) was inspired by , harmonic drive is the only thing that gave me the desired hand positioning accuracy.Then everything died out for me after I received 6 pieces of these harmonic drives (each has a price of 1k) after 6 months and when I began to put everything together, it turned out that I should have taken a few others and then I gave up, maybe the desire to wait for another 6 the month for the next iteration was gone, and the inner perfectionist told me that he didn’t want to sculpt crutches with what he had.So here is my question, the answer to which I could not find then: - since my goal was to make a collaborative robot, it was very important to read the torque from the motors so that the hand would stop when it collided with a person, I found the torque sensors separately, but they were prohibitively expensive, especially if you put them on all links, I know that Universal Robots somehow read torque from motors, but I don’t understand how this is possible, given that after the motor there is a harmonic drive with a gear ratio of 1:100 those its just not scrolling backwards how do you do it?And finally, you got out with the motors, by the way, I had no problems with their delivery, but what about the harmonic drive?While waiting for the delivery, I assembled a cart for myself with such a good payload to study spatial orientation algorithms, but most of all it was interesting to play with the kinect, of course, here is the first unennobled version: https://www.youtube.com/watch?v=4DSkNnre7IIThanks for the comment, Denis!You're right - UR sets the pace and everyone looks back at it one way or another.Maxons are good motors.We looked in the direction of RoboDrive, but did not grow together.As for the gearbox, we use the harmonic drive gearbox in our servos.This is one of the few components we purchase.True, now we tried other manufacturers.In particular, the American ConeDrive.So far, there are no results of long-term tests, I think that in the near future there will be sufficient operating time and it will be possible to draw a conclusion.As for collaboration, you are right.Universal Robots does not use torque sensors.They cost, for example, in Kuka, but the price of Kuka is 2-3 times higher than the price of UR (depending on the model).We measure the amount of current drawn to determine the collision.There is a certain software algorithm that detects abnormal current changes and at that moment stops the cobot.In practice, it turns out that a person feels a small push.This is where fine tuning is needed.With a competent approach, this is an effective and inexpensive option for collaboration.Thanks for the answer, now I was sitting and trying to remember what controllers I tried then: 1) so I started everything with controllers from https://www.roboteq.com/index.php/roboteq-products-and-services/brushless-dc -motor-controllers, the prices there were absolutely ridiculous and therefore I started with them, but something didn’t suit me, now I won’t even say that it was so long ago, but I tried one of the cheapest models, it was worth taking more mb 2 ) then I tried this one in my opinion http://www.galilmc.com/motion-controllers/multi-axis/dmc-41x3, but after playing around it was decided to change the design, I didn’t want to pull a bunch of reasons through my whole arm to an external controller, it was decided it was to try to make a controller right next to the engine, for the same reason I discarded the controllers from Maxon, although by that time just one had already come to me for testing, but I didn’t even turn it on and it’s in the box because the dimensions didn’t allow it to be built into hand 3) in the end, I settled on controllers from https://www.amc.com/ Something like this.in fact, the problem of determining torque is both very simple and very complex at the same time.Simple, because the current consumed by the motor is directly proportional to the force that the motor creates and is determined by the constant Kt (Torque = Kt * Current).This constant is different for each motor and is determined either by the manufacturer's datasheet (as in the case of Maxon), or, preferably, experimentally.Accordingly, knowing how the robot is located, how much each joint of the robot "weighs", where the gravity vector is directed (the robot is standing on the floor or wall) and what load is in the grip of the robot, you can calculate how much current each robot motor should consume.Accordingly, if the motor consumes more or less, then the robot has encountered an obstacle.The task is difficult, because there are a lot of nuances how to filter the noisy current value, how to determine this emergency collision or a person just interacting with the robot, etc.To understand how a robot works, I recommend reading either "Introduction to Robotics: Mechanics and Control" by John Craig (in Russian translation "Introduction to Robotics. Mechanics and Control") or "Robotics, Vision and Control: Fundamental Algorithms In MATLAB" by Peter Corke.The last book has a library available for download in matlab, on which you can test all the robot control algorithms.After reading these two books, I am sure that there will be no fewer questions, but they will become more targeted.well, let's not lie - sales have just started).In April, we make the first shipments on pre-orders.Among those who pre-ordered - manufacturers.Food products, shoe production, assembly production of parts, medical topics, metalworking.These are key.Among those who are interested and think about how to attach to the project are a bank, startups where a manipulator is needed, a fun sector - a barista robot, etc.Basically, they take it for 2 reasons.the first time a person fails.Those.a large amount of routine (food - packaging and pre-sorting with a large assortment).the second is to prevent marriage (shoe production).a person may have something to "itch", he may think, gape, make a mistake.do not "according to the guest".The main goal is to automate production.to do more, qualitatively and without marriage.We presented our cobot to Sberbank 2 weeks ago.And in the process of communication, their representative says, "we need to make a robot barista."Why, I say, a jar of a robot barista?!Answer: EVERYONE needs a robot barista.No show anywhere.At the moment, one model is on sale - a load capacity of 3 kg, a working radius of 700 mm.Detailed datasheet can be downloaded here: http://bit.ly/PULSE_Datasheet_Rus Cost: 18.500 USD.Good afternoon, Igor.1. Source of funding - the founder's own funds.Now - 25 people are working.We receive a salary) 2. Delivery through services such as UPS, etc. When delivering to the Russian Federation, there is no need to clear customs.For deliveries outside the Customs Union, customs agents help.Our sales manager can tell you more details (write to me on FB) 3. Cycloidal gearboxes did not fit in size and gear ratio.Plus they are noisier.I think, as we get stronger, we will be able to take on our gearbox / Because so far this is the bottleneck.